A large quantity of streaming media items, such as audio items and video items, are available to users through cable television, Internet Protocol television (IPTV), internet radio and other distribution systems. Recommendation systems have been proposed to help users find and play those items that may be of interest to them, and to filter out items that may not be of interest to them. For television programming, some recommendation systems cluster people based on what shows they enjoy. However, different people may enjoy the same show for different reasons. For example, a first group of people may like a television show such as “Law & Order” because of its being an investigative drama, and a second group of people may like the same television show because of its being a court drama. Using a current voting system for television programming, both the first group and the second group would indicate that they enjoy this television show. However, current recommendation systems may use this information to make an ill-advised recommendation of other court dramas to the first group, and/or other investigative shows to the second group.